Learning spatial referential words with mobile robots Simon Dobnik
This event took place on 8th March 2006 at 9:00am (09:00 GMT)
Knowledge Media Institute, Berrill Building, The Open University, Milton Keynes, United Kingdom, MK7 6AA
Natural language contains a number of word categories that are referential in nature. Their full semantics can only be evaluated by examining the context in which the words are used. An important and challenging group of words are those that describe space. A system has been developed where the meanings of spatial words such as near, left and behind are learnt by a mobile robot from its experience of environment: by abstracting over the properties of its knowledge of environment and the descriptions that a human commentator used to describe it. Learning is performed offline using statistical and symbolic learning techniques. With the knowledge that it acquired the robot is able to generate new descriptions of new environments. Users can query the robot through a simple dialogue interface using spoken natural language.
Natural language contains a number of word categories that are referential in nature. Their full semantics can only be evaluated by examining the context in which the words are used. An important and challenging group of words are those that describe space. A system has been developed where the meanings of spatial words such as near, left and behind are learnt by a mobile robot from its experience of environment: by abstracting over the properties of its knowledge of environment and the descriptions that a human commentator used to describe it. Learning is performed offline using statistical and symbolic learning techniques. With the knowledge that it acquired the robot is able to generate new descriptions of new environments. Users can query the robot through a simple dialogue interface using spoken natural language.
Click below to play the event (31 minutes)